set silence trimming threshold in config

config.json

@@ -24,6 +24,7 @@
         "mel_fmin": 0.0,         // minimum freq level for mel-spec. ~50 for male and ~95 for female voices. Tune for dataset!!
         "mel_fmax": 8000.0,        // maximum freq level for mel-spec. Tune for dataset!!
         "do_trim_silence": true  // enable trimming of slience of audio as you load it. LJspeech (false), TWEB (false), Nancy (true)
+        "trim_db": 60,          // threshold for timming silence. Set this according to your dataset.
     },
 
     // DISTRIBUTED TRAINING

utils/audio.py

@@ -24,6 +24,7 @@ class AudioProcessor(object):
                  clip_norm=True,
                  griffin_lim_iters=None,
                  do_trim_silence=False,
+                 trim_db=60,
                  sound_norm=False,
                  **_):
 
@@ -46,6 +47,7 @@ class AudioProcessor(object):
         self.max_norm = 1.0 if max_norm is None else float(max_norm)
         self.clip_norm = clip_norm
         self.do_trim_silence = do_trim_silence
+        self.trim_db = trim_db
         self.sound_norm = sound_norm
         self.n_fft, self.hop_length, self.win_length = self._stft_parameters()
         assert min_level_db != 0.0, " [!] min_level_db is 0"
@@ -217,7 +219,7 @@ class AudioProcessor(object):
         margin = int(self.sample_rate * 0.01)
         wav = wav[margin:-margin]
         return librosa.effects.trim(
-            wav, top_db=40, frame_length=self.win_length, hop_length=self.hop_length)[0]
+            wav, top_db=self.trim_db, frame_length=self.win_length, hop_length=self.hop_length)[0]
 
     @staticmethod
     def mulaw_encode(wav, qc):